Course contentsShow
← Data Science
Lesson 710 of 2,14519. Time Series FoundationsFree lesson

Additive vs Multiplicative Models

Choose between additive decomposition (components sum) and multiplicative (components multiply) based on data behavior.

Additive vs Multiplicative Models

You'll learn how to choose the right decomposition structure based on whether your time series components add together or multiply together.

The Core Difference

When decomposing a time series into Trend, Seasonality, and Irregular components, you face a fundamental choice: do these pieces combine by adding or by multiplying?

Additive model: Observed = Trend + Seasonality + Irregular

Use this when seasonal fluctuations stay roughly the same size regardless of the trend level. If your data increases from 100 to 1000 units but the seasonal "swing" remains ±10 units throughout, that's additive behavior.

Multiplicative model: Observed = Trend × Seasonality × Irregular

Use this when seasonal fluctuations grow (or shrink) proportionally with the trend. If sales triple and the seasonal swings also triple, that's multiplicative behavior. This is extremely common in business and economic data where percentage changes matter more than absolute amounts.

How to Decide

Look at your plot:

  • Constant-width seasonal swings → additive
  • Widening or narrowing seasonal swings → multiplicative

Think about the real-world process:

  • Measuring temperature differences? Probably additive (±5°C stays ±5°C)
  • Measuring sales growth? Probably multiplicative (seasonal boosts are percentage-based)

A Helpful Analogy

Imagine a ladder (trend) with decorations (seasonality). In an additive world, each decoration adds the same fixed height. In a multiplicative world, each decoration stretches proportionally—higher rungs get proportionally bigger decorations.

Key Takeaway: Choose additive decomposition when seasonal variations remain constant in absolute terms; choose multiplicative when they scale proportionally with the trend level.